Introduction
Priya Raman still remembers the first X-ray where she saw the sign too late. It was not a dramatic failure, more a small delay that stayed in her mind. Dentistry, she learned, is full of such narrow margins between routine and risk.
Priya lives in Lausanne and leads a small health-AI team. She came from clinical work, not from a computer science lab, and that makes her suspicious of systems that sound more certain than a careful dentist would.
Story of the Path into AI
A hand injury limited Priya’s work in the practice and forced a professional turn. She began asking whether AI could help colleagues notice routine findings earlier without shifting medical responsibility to software. Investors wanted speed; clinics wanted safety. Priya had to learn enough programming to ask better questions and enough validation to resist easy claims.
She worked with computer scientists, dental practitioners and patient advocates. Her first model marked abnormalities in dental X-rays and displayed uncertainty. It made several useful suggestions and one alarming mistake: a shadow from positioning was highlighted as suspicious. Priya used that case to redesign the interface. The model was not allowed to speak like a diagnosis.
Current Work
Today Priya’s team develops assistance systems for practices and tests them clinically. In pilot settings, suggestions appear as a second opinion. If the system sounds too confident, the wording is changed. Documentation matters: what did the model mark, what did the dentist decide, and why?
The early feedback is practical rather than glamorous. Some practices report more structured documentation and fewer missed routine checks. Priya avoids claims of automatic diagnosis. In her view, the real product is a more attentive workflow, not a machine that pretends to be a dentist.
Personal Advice
“In medicine, a cautious assistant is worth more than a loud prophet,” Priya says. Her advice to clinical founders is to love validation more than demos. A tool that cannot survive careful testing does not belong near patients.
Key Facts
Age and place: 36, Lausanne.
Background: dentistry, career change after injury, clinical responsibility.
Entry into AI: model for marking abnormalities in dental X-rays with uncertainty.
Focus today: health AI for clinical assistance.
Typical tools: image analysis, clinical validation, human-in-the-loop design.
Werkstattnotiz
Priya keeps the misread shadow in every training deck. It is not there to embarrass the model, but to discipline the room. The image asks a simple question: if this were your patient, how much certainty would be enough before you changed the plan?